Absolute threshold

The absolute threshold can be compared to the difference threshold, which is the measure of how different two stimuli must be for the subject to notice that they are not the same.

Vision

A landmark 1942 experiment by Hecht, Shlaer, and Pirenne assessed the absolute threshold for vision. They tried to measure the minimum number of photons the human eye can detect 60% of the time, using the following controls:

1. Dark adaptation – the participants were completely dark adapted (a process lasting forty minutes) to optimise their visual sensitivity.
2. Location – the stimulus was presented to an area of the right eye where there is a high density of rod cells, 20 degrees to the left of the point of focus (i.e. 20 degrees to the right of the fovea). Roughly this degree of eccentricity (about 20 degrees) has the highest rod density across the whole retina. However, the corresponding location on the right retina, 20 degrees to the left, is very near the blind spot.
3. Stimulus size – the stimulus had a diameter of 10 minutes of arc (1 minute = 1/60th of a degree). Although not explicitly mentioned in the original research paper, this ensured that the light stimulus fell only on rod cells connected to the same nerve fibre (this is called the area of spatial summation).
4. Wavelength – the stimulus wavelength matched the maximum sensitivity of rod cells (510 nm).
5. Stimulus duration – 0.001 second (1 ms).

The researchers found that the emission of only 5–14 photons could elicit visual experience. However, only about half of these entered the retina, due to reflection (from the cornea), absorption, and other factors relating to transmittance of the ocular media. The researchers estimated that 5 to 14 of the estimated 500 rods in the test area would each absorb one photon, with a 4% chance that one rod would absorb two photons.

A second absolute threshold for vision involves the minimum photon flux (photons per second per unit area). In this case the light covers a wide field over an extended period of time instead of being concentrated on one spot on the retina in a short burst. Knowing the pupil diameter and the wavelength of the light, the result can be described in terms of luminance (~0.000001 candela per square meter or 10⁻⁶ cd/m²) or retinal illuminance (~0.00002 Trolands). By including estimates for the probability of an average photon being absorbed by an average rod cell, the threshold stimulation for the rods is approximately one photon absorption per second per 5000 rods.

In terms of total absolute power sensitivity, Denton and Pirenne in The Journal of Physiology in 1954 found that for diffuse, extended sources i.e. a relatively large (~ 45-degree wide source, as viewed by the subject) frosted glass aperture, and a long (5 second) observation and decision time, the human eye could begin to reliably distinguish the lit from unlit glass at a power level of approximately 7.6 × 10⁻¹⁴ watts / steradian-cm² at the eye for green (510 nm) light. This power level was dependent on the wavelength of the light used according to the usual luminosity curve. For white light, the absolute sensitivity found was 5.9 × 10⁻¹⁴ watts / steradian-cm². This base sensitivity varied only about 0.03 log steps between monocular (one-eyed) or binocular (two-eyed) vision.

In 1972 Sakitt conducted an experiment that combined elements of signal detection and threshold theory. Two key elements of the study were a high tolerance for false positives and a multiple-choice option on deciding whether or not a light was seen. In the classic studies described above, the tolerance for false positives was so low that threshold was biased upward. Based on statistical analysis of a large number of trials, 6 photons each absorbed by one rod near-simultaneously looked "very bright", 5 photons looked "bright", 4 photons "a moderate light", 3 photons "a dim light". Two observers were able to see 2 photons as "slightly doubtful if a light was seen." One observer saw a single photon as "very doubtful if a light was seen." Zero photons were seen as "did not see anything".